Abstract -------- The aim of the study was to determine the factors affecting teacher’s judgment and its accuracy while estimating performance in tests of cognitive abilities and academic achievement of students, and on which student abilities the evaluation was based on. The study included 223 students (38% females) of the fourth grade who were administered the test of cognitive abilities CFT 20-R and the test of academic (mathematical) achievement TIM3–5. Their performance was estimated by 10 teachers (8 females) using a 9-point scale and percentile. Furthermore, for a part of the children (n = 130), 6 of these teachers determined the probability with which the child would correctly solve two particular items of test TIM3–5. The most accurate method of estimation is the percentile rating, both for the cognitive abilities (r = .56) and the academic achievement (r = .63). The length of teacher experience and the level of evaluated student ability has no effect on the accuracy of judgment. The student’s gender has no effect on the accuracy of estimation or the level of ascribed ability. Teachers tended to underestimate students in a moderately difficult item of test TIM3–5, and vice versa in case of a difficult item. The study further presents a theoretical model in which the teacher’s impression of the student is based on the student’s cognitive abilities and academic achievement. Empirical verification of this model has shown that the basis for teacher’s judgment is primarily the student’s intelligence, which is manifested indirectly in the student’s academic achievement. File structure ---------- **analytical script.R** - main script for all the analyses performed in the paper. To run properly, keepbe sure to have an *export* folder within the working directory. The R version and versions of all the packages are quoted. **Data_judgement.xlsx** - all the data on the first sheet, codebook on the second. **/export** folder - table with all the objects exported from R using an *analytical script.R* file.